Seeing difference in SUBTASK EFFECTIVENESS between machines

Products

IDMS IDMS - Database IDMS - ADS

Issue/Introduction

There is a difference in CPU effectiveness for the same IDMS region between two machines.

Question is -- is this normal? Does it have something to do with different architecture?

D SUBTAS EFF

*** Subtask display ***

Subtask % CPU

Name Elapsed time Total CPU time effectiveness

-------- -------------- -------------- -------------

MAINTASK 00:00:14.9540 00:00:00.4021 02

SUBT0001 00:02:09.7507 00:00:21.9015 16

SUBT0002 00:34:21.3877 00:22:11.5570 64

-------- -------------- -------------- -------------

Totals 00:36:46.0924 00:22:33.8606 61

      D STAT SYS
12:58:29.71 20/077 Current Time       00:19:05.93 Tot Sys Time
13:19:11.24 20/075 Startup Time       00:03:26.82 Tot User Time

TASKS:         66475 Processed        243 Abended               47 Max Tasks
                4782 System             0 Runaway                0 Times At Max
                   8 Deadlocks          6 Dead Victims

TRANS:        151422 Processed     108283 Norm Cmp   103 Max Conc
                6229 Ext Proc        6179 Ext Norm     8 Ext Conc    10 Max Erus
                   0 Dist Proc          0 Dist Norm    0 Dist Conc

DATABASE:   35342552 Calls      514547495 Pages Rqst     555031959 Recs Rqst
                   0 Buff Wait   34226065 Pages Read     123495910 Recs Cur R/U
             2245440 Page Writ     169096 Calc Noflo        209328 Via Noflo
           672830765 Tot Locks       2061 Calc Ovflo         57925 Via Ovflo
                                     6866 Frag Stord             0 Recs Reloc
   INDEX:      16582 SR8 Splits     15907 SR8 Stores            42 SR7 Stores
                 143 SR8 Spawns     15527 SR8 Erases            42 SR7 Erases
                2821 Orph Adopt   2352527 Ix Searches            1 Min Level
                                  8027787 Lvls Srchd             7 Max Level
     SQL:    3797982 Commands     6268769 Tupls Fetched     169868 Rows Inserted
                   0 AM Recomp     446122 Rows Updated      173222 Rows Deleted
                                    95120 Sorts                  1 Sort Min
                                  1273784 Tuples Sorted      36972 Sort Max

JOURNAL:                    0 Buff Waits                0 User Putjrnl
Page   376073 0-10   24870 11-20   20546 21-30   80161 31-40   16331 41-50
Dist    15818 51-60   14333 61-70   25830 71-80 107022 81-90 949999 91-100

INTERNAL:       RLEs    RCEs    DPEs   Stack
                5596    5538     681     967 HWM
                8000    8000    1100    3200 Sysgen Threshold
                   0       0       0         Times Exceeded

STORAGE:     8604995 Gets         8518240 Frees            Gets for type
                   0 PGFIXs             0 PGFREEs           3266699 DB
                   0 Pages Fxd          0 Pages Freed           100 SHK
             7601777 Scan 1             0 PGRLSEs                 1 SHR
             1003218 Scan 2             0 Pages Relsd       1271055 SYS
                                                              64301 USK
                   0 SOS COUNT                              4002839 USR

PROGRAM:             Act Loads Pages Load Wait/Space
          Non-Reent          3           8           0
              Reent         40         962           0
       XA Non-Reent        110        1042           0
           XA Reent      11148     4306746           0

SCRATCH:      286550 Gets     293125 Puts     292860 Dels
QUEUE:       28718 Gets      27146 Puts       4824 Dels         34 Task AutoSt

TIME:        4705091 Gets     180327 Post         84 Started Tasks
                3859 Wait     180411 Canc

USERS:            66 Signed on                    95 HWM

Environment

Release : 16.0

Component : CA ADS

Resolution

The lower subtask effectiveness on the enterprise class machine could be normal. It means IDMS is not getting the amount of CPU it needs and this doesn't necessarily mean that it will be unresponsive with slow performance.

CPU Effectiveness represents the amount of CPU time an IDMS CV needed
vs. the amount of CPU time it actually received. The number should be as
close to 100 percent as possible. Then again, a CPU Effectiveness value
of 20 percent could also be considered acceptable if you are happy with
performance and meeting service level agreements.

The value is derived as follows:

When the IDMS system is dispatched by the operating system one of the
very first things we do is issue a STCK instruction. This represents the
time we were dispatched. When IDMS has finished all it's processing we
again issue a STCK instruction. This gives us our ending time. The
difference between the two is the amount of time the IDMS system was
dispatched.

Between the two STCKs above, the IDMS system is calculating the amount
of CPU time it is actually getting using operating systems related
timing functions. While the IDMS system is up and dispatched it can be
interrupted due to Paging, Swapping, or Higher priority work. If such an
interruption occurs the CPUTIME used during this interrupt is not
attributed to IDMS.

Example:
Assume the IDMS system is dispatched at STCK time 0 and goes to sleep at
STCK time 10. During these 10 seconds another application running at a
higher priority is dispatched. Because this other application is running
at a higher priority the CPU is taken from the IDMS system. This other
application runs for 3 seconds then gives up control. The IDMS system
which had been interrupted is then allowed to continue.

IDMS was dispatched for 10 seconds. Higher priority work interrupted for
3 seconds. The CPU effectiveness is 70 percent.